You are on page 1of 27

RF engineering basic concepts: S-parameters

F. Caspers
CERN, Geneva, Switzerland

Abstract
The concept of describing RF circuits in terms of waves is discussed and the
S-matrix and related matrices are defined. The signal flow graph (SFG) is
introduced as a graphical means to visualize how waves propagate in an RF
network. The properties of the most relevant passive RF devices (hybrids,
couplers, non-reciprocal elements, etc.) are delineated and the corresponding
S-parameters are given. For microwave integrated circuits (MICs) planar
transmission lines such as the microstrip line have become very important.

1 S-parameters
The abbreviation S has been derived from the word scattering. For high frequencies, it is convenient
to describe a given network in terms of waves rather than voltages or currents. This permits an easier
definition of reference planes. For practical reasons, the description in terms of in- and outgoing
waves has been introduced. Now, a 4-pole network becomes a 2-port and a 2n-pole becomes an n-
port. In the case of an odd pole number (e.g., 3-pole), a common reference point may be chosen,
attributing one pole equally to two ports. Then a 3-pole is converted into a (3+1) pole corresponding
to a 2-port. As a general conversion rule, for an odd pole number one more pole is added.

I1 I2

Fig. 1: Example for a 2-port network: a series impedance Z

Let us start by considering a simple 2-port network consisting of a single impedance Z
connected in series (Fig. 1). The generator and load impedances are ZG and ZL, respectively. If Z = 0
and ZL = ZG (for real ZG) we have a matched load, i.e., maximum available power goes into the load
and U1 = U2 = U0/2. Note that all the voltages and currents are peak values. The lines connecting the
different elements are supposed to have zero electrical length. Connections with a finite electrical
length are drawn as double lines or as heavy lines. Now we would like to relate U0, U1 and U2 to a
and b.

67

F. C ASPERS

1.1 Definition of ‘power waves’
The waves going towards the n-port are a = (a1, a2, ..., an), the waves travelling away from the n-port
are b = (b1, b2, ..., bn). By definition currents going into the n-port are counted positively and currents
flowing out of the n-port negatively. The wave a1 going into the n-port at port 1 is derived from the
voltage wave going into a matched load.

normalized to √Z0. Z0 is in general an arbitrary reference impedance, but usually the characteristic
In order to make the definitions consistent with the conservation of energy, the voltage is

impedance of a line (e.g., Z0 = 50 Ω) is used and very often ZG = ZL = Z0. In the following we assume
Z0 to be real. The definitions of the waves a1 and b1 are

incident voltage wave ( port 1)
= = =
U0 U1inc
a1
2 Z0 Z0 Z0

reflected voltage wave ( port 1)
(1.1)

=b1 =
U1refl
Z0 Z0

Note that a and b have the dimension power [1].

The power travelling towards port 1, P1inc, is simply the available power from the source, while
the power coming out of port 1, P1refl, is given by the reflected voltage wave.
2 2

= = =
U1inc I1inc
inc 1 2
P1 a1 Z0
2 2Z0 2
(1.2)
2

= = =
1 2 U1refl I1refl
refl
P1 b1 Z0
2 2 Z0 2
Note the factor 2 in the denominator, which comes from the definition of the voltages and
currents as peak values (‘European definition’). In the ‘US definition’ effective values are used and
the factor 2 is not present, so for power calculations it is important to check how the voltages are
defined. For most applications, this difference does not play a role since ratios of waves are used.
In the case of a mismatched load ZL there will be some power reflected towards (!) the 2-port
from ZL, leading to some incident power at port 2.

P2inc =
1 2
a2 . (1.3)
2
There is also the outgoing wave of port 2 which may be considered as the superimposition of a wave
that has gone through the 2-port from the generator and a reflected part from the mismatched load.
= =
inc
We have defined a1 U 0 / (2 Z 0 ) U inc / Z 0 with the incident voltage wave U . In analogy to that
we can also quote for the power wave a1 = I inc Z 0 with the incident current wave Iinc. We obtain the
general definition of the waves ai travelling into and bi travelling out of an n-port:

68

n) can be written as a system of n linear equations (ai being the independent variable. Using Eqs.PARAMETERS Ui + Ii Z0 ai = 2 Z0 Ui − Ii Z0 bi = (1.6) ( ) 2 = ai a*i − bi bi* 1 Pi 2 The term (ai*bi – aibi*) is a purely imaginary number and vanishes when the real part is taken. Ui and Ii can be obtained for a given ai and bi as Ui = Z 0 ( ai + bi ) = U iinc + U irefl ( ai −= bi ) (1. S21 is the forward transmission (from port 1 to port 2).4) and (1. Z0 Z0 For a harmonic excitation u(t) = Re{U e jωt } the power going into port i is given by Pi = 1 { } Re U i I i* {( ) ( )} 2 = Re ai a*i − bi bi* + a*i bi − ai bi* 1 Pi (1. (1. 1. in matrix formulation b=Sa (1.. bi the dependent variable) b1 = S11a1 + S12 a2 b2 = S21a1 + S22 a2 (1. including the port connected to the generator (matched generator).4) 2 Z0 Solving these two equations.7) or. When measuring the S-parameter of an n-port.7) we find the reflection coefficient of a single impedance ZL connected ( Z / Z0 ) − 1 to a generator of source impedance Z0 (Fig.2 The S-matrix The relation between ai and bi (i = l. S12 the reverse transmission (from port 2 to port 1) and S22 the output reflection coefficient.8) The physical meaning of S11 is the input reflection coefficient with the output of the network terminated by a matched load (a2 = 0). 1. case ZG = Z0 and Z = 0): U −I Z Z − Z0 S11 = 1 = 1 1 0 = L =Γ= L ( Z L / Z0 ) + 1 b a1 a = 0 U1 + I1Z 0 Z L + Z 0 (1.. all n ports must be terminated by a matched load (not necessarily equal value for all ports).5) = 1 U irefl Ii .RF ENGINEERING BASIC CONCEPTS : S. 69 .9) 2 which is the familiar formula for the reflection coefficient Γ (sometimes also denoted as ).

While the S-matrix exists for any 2-port.3 The transfer matrix The S-matrix introduced in the previous section is a very convenient way to describe an n-port in terms of waves. in general due to symmetries etc. active components such as amplifiers are generally non-reciprocal. no transmission between port 1 and port 2. F.g.).10) and in a similar fashion we get b2 U 2 − I 2 Z 0 S= = = 2Z0 a1 U1 + I1Z 0 2 Z 0 + Z 21 . the number of independent coefficients is much smaller. the T-matrix is not defined. A very straightforward manner for the problem is possible with the T-matrix (transfer matrix). assuming again ZG = ZL = Z0. Please note that for this case we obtain S11 + S21 = 1. 3]): TM = T1T2  Tm . capacitors.11) Due to the symmetry of the element S22 = S22 and S12 = S21. [2. plasmas etc. • An n-port is reciprocal when Sij = Sji for all i and j.  Z0 + Z  0 (1.13) The conversion formulae between the S and T matrix are given in Appendix I.12)  2Z + Z 2 Z 0 + Z  Z  0 1. transformers. It is very well adapted to measurements. e. However. Most passive components are reciprocal (resistors.. The T-matrix TM of m cascaded 2-ports is given by (as in Refs. The full S-matrix of the element is then  Z0 + Z   2Z + Z 2Z0 + Z  Z S= . U2 = U0 . (1. While the Sij may be all independent.. From the definition of S11 we have b1 U 1 − I 1 Z 0 S 11 = = a1 U 1 + I 1 Z 0 Z0 + Z U1 = U 0 . I1 = = −I 2 Z0 U0 2Z 0 + Z 2Z 0 + Z 2Z 0 + Z ⇒ S 11 = Z 2Z 0 + Z (1.  1   21 T22   b2  (1. it is not well suited for characterizing the response of a number of cascaded 2-ports. which directly relates the waves on the input and on the output [2]  b1   T11 T12   a2   a  = T   . C ASPERS Let us now determine the S-parameters of the impedance Z in Fig. 1. (1. 2 Properties of the S-matrix of an n-port A generalized n-port has n2 scattering coefficients.14) Note that in the literature different definitions of the T-matrix can be found and the individual matrix elements depend on the definition used. 70 . in certain cases. except for structures involving magnetized ferrites. etc.

2) with the first of Eq.e. in case of lossless networks. S12 S21 (2..1) which yields three conditions S11 + S 21 =1 2 2 + S 22 =1 2 2 S12 (2.4) − arg S11 + arg− S12 = + S22 π + S21 arg arg where arg(x) is the argument (angle) of the complex variable x.3) Splitting up the last equation in the modulus and argument yields S11 S12 = S21 S22 and (2. For a two-port this means  S*  S11 S12   1 0  S =  11  =  *  S 22   0 1  * * T S 21  S12 * S 22  S 21 (2. 2. S†S = 1. α the line attenuation in [neper/m] and β = 2π/λ with the wavelength λ. 71 . • An N-port is passive and lossless if its S matrix is unitary. where x† = (x*)T is the (S ) conjugate transpose of x. (2. i. Combining Eq. Their phases change when the reference plane is moved.RF ENGINEERING BASIC CONCEPTS : S.PARAMETERS • A two-port is symmetric.5) S11= 1 − S12 2 . In general the S-parameters are complex and frequency dependent. (2. Thus any lossless 2-port can be characterized by one modulus and three angles. Often the S-parameters can be determined from considering symmetries and.4) then gives = =S11 S22 . For a lossless line we obtain |S21| = 1.2) * S11 S12 + S*21S22 = 0 (2. energy conservation. when it is reciprocal (S21 = S12) and when the input and output reflection coefficients are equal (S22 = S11).1 Examples of S-matrices 1-port • Ideal short S11 = −1 • Ideal termination S11 = 0 • Active termination (reflection amplifier) S11 > 1 2-port • Ideal transmission line of length l  0 e −γ l  S =  −γ l  e 0  where γ = a + jβ is the complex propagation constant.

while for the gyrator ϕ12 = ϕ21 + π. The following three components have two of the above characteristics: • Resistive power divider: It consists of a resistor network and is reciprocal. e. 50 Ω. to avoid reflections from a load back to the generator. An attenuator can be realized. however. for example. passive gyrators can be realized using magnetically saturated ferrite elements. in the microwave range. Gyrators are often implemented using active electronic components. It can be realized with three resistors in a triangle configuration. • Ideal amplifier  0 0 S =   G 0 with the gain G > 1. reciprocal. but it is not reciprocal. An ideal gyrator is lossless (S†S = 1). matched at all ports but lossy. 3-port Several types of 3-port are in use. The attenuation in decibel is given by A = -20*log10(S21). • Ideal isolator  0 0 S =    1 0 The isolator allows transmission in one direction only.g.. for example. C ASPERS • Ideal phase shifter  0  − jφ 12 S =  − jφ   e 21 0  e  For a reciprocal phase shifter ϕ12 = ϕ21. 1 Np = 8. The values of the required resistors are k −1 R1 = Z 0 k +1 R1 R1 Port 1 R2 Port 2 R2 = Z 0 2 2k k −1 where k is the voltage attenuation factor and Z0 the reference impedance.. 72 . and matched at all three ports at the same time. circulators. the resulting circuit is similar to a 2-port attenuator but no longer matched at port 1 and port 2.686 dB. T junctions. with three resistors in a T circuit. it is used. With port 3 connected to ground. It can be shown that a 3-port cannot be lossless.g. e. • Ideal. F. reciprocal attenuator  0 e −α  S =  −α  e 0  with the attenuation α in neper. power dividers.

Fig.RF ENGINEERING BASIC CONCEPTS : S. matched at all ports. The magnetic properties of a 73 .plane splitters 1 −1 2 1 2 1  1  . 2 1 2 1 3 Fig. 3). A circulator.and E. 1 =  −1 =2 S E − 1 2  2  2  SH 1 1  2 0   2 − 2 0   2  • The ideal circulator is lossless.  0 1 0   When port 3 of the circulator is terminated with a matched load we get a two-port called isolator. The real and imaginary part of each normally magnetized into saturation by an external magnetic field. the S-matrix of the isolator has the following form: 0 0 1   S = 1 0 0 . Accordingly. which lets power pass only from port 1 to port 2 (see section about 2-ports). for appropriate reference planes for H. 3: 3-port circulator and 2-port isolator. 2: The two versions of the H10 waveguide T splitter: H-plane and E-plane splitter Using the fact that the 3-port is lossless and there are symmetry considerations one finds. like the gyrator and other passive non-reciprocal elements contains a volume of ferrite. This ferrite is saturated RF ferrite have to be characterized by a µ-tensor. but not reciprocal. The circulator can be converted into isolator by connecting a matched load to port 3. A signal entering the ideal circulator at one port is transmitted exclusively to the next port in the sense of the arrow (Fig.PARAMETERS  0 1   2 1 S=1 1  2 Port 1 Z0/3 Z0/3 Port 2  2 2 0 1 0  Z0/3  2  1 2 Port 3 • The T splitter is reciprocal and lossless but not matched at all ports.

The magnetically polarized ferrite provides the required nonreciprocal properties. As a result. 4). The µ+ and µ- (Fig.and left-hand circular polarized wave traversing the ferrite depicted denoted as µ′+ and µ′. 5 and 6 practical implementations of circulators are shown. At the gyromagnetic resonance the right-hand right. = 28 GHz/T (gyromagnetic ratio) = 28 GHz/T (gyromagnetic ratio) Fig. 74 .and left-hand circularly polarized waves in a microwave ferrite are polarized has high losses. C ASPERS complex element µ are µ′ and µ′′. 4]. In this figure the µ′ and the µ′′ for the right. They are strongly dependent on the bias field.and left-hand circular polarized waves are represent the permeability seen by a right. 4: Real part µr′ and imaginary part µr'' of the complex permeability µ. power is only transmitted from port 1 to port 2. as can be seen from the peak in the lower image. F. In Figs. from port 2 to port 3 and from port 3 to port 1. The µ+ and µ. A detailed discussion of the different working principles of circulators can be found in the literature [2.respectively.for the real part of µ and correspondingly for the imaginary part.

Therefore the Faraday isolator allows transmission only from port 1 to port 2.PARAMETERS port 3 port 1 ground planes ferrite disc port 2 Fig. After a transition to a circular waveguide.. reciprocal elements. Then follows a transition to another rectangular waveguide which is rotated by 45o such that the rotated forward wave can pass unhindered. 8). The horizontal attenuation foils dampen this mode. while they hardly affect the forward wave. 6: Stripline circulator The Faraday rotation isolator uses the TE10 mode in a rectangular waveguide. 8: Hybrid ‘T’. non-reciprocal elements extends from about 50 MHz up to optical wavelengths (Faraday rotator) [4]. 180° hybrid. 5: Waveguide circulator Fig.RF ENGINEERING BASIC CONCEPTS : S. which has a vertically polarized H field in the waveguide on the left (Fig. Finally. However. 7). it should be noted that all non-reciprocal elements can be made from a combination of an ideal gyrator (non-reciprocal phase shifter) and other passive.g. 4-port T-hybrids or magic tees. e. Ideally there is no crosstalk between port 3 and port 4 nor between port 1 and port 2 75 . a wave coming from the other side will have its polarization rotated by 45o clockwise as seen from the right side. The S-matrix of a 4-port As a first example let us consider a combination of E-plane and H-plane waveguide ‘T’s (Fig. Magic ‘T’. Fig. 7: Faraday rotation isolator The frequency range of ferrite-based. In the waveguide on the left the backward wave arrives with a horizontal polarization. the polarization of the waveguide mode is rotated counter clockwise by 45o by a ferrite. attenuation foils Fig.

respectively. Today. This is shown in Fig. With a suitable choice of the reference planes the very simple S-matrix given above results. The signal from generators connected to the collinear arms of the magic T. but as coaxial lines and printed circuits. In a simple vertical-loop pickup the signal outputs of the upper and lower electrodes are connected to arm 1 and arm 2. a 4-port may be lossless. (b. 9: Waveguide directional couplers: (a) single-hole. microstrip). A selection of possible waveguide couplers is depicted in Fig. Contrary to a 3-port. Another important element is the directional coupler. and (d. when there is no equal propagation in the two directions. The signal from generator 1 is split and fed signals cancel at the ∆ port and the sum signal shows up at the Σ port. certain measures are required to make the ‘T’ a ‘magic’ one. and matched at all ports simultaneously. a b c λg/4 d e Fig. the magic ‘T’ is around one octave or the equivalent H10-mode waveguide band.. In practice. The coupler is directional when the coupled energy mainly propagates in a single travelling wave. and coupling is adjusted such that part of the power linked to a travelling wave in line 1 excites travelling waves in line 2.e. F. Broadband versions of 180° hybrids may have a frequency range from a few MHz to some GHz. C ASPERS This configuration is called a Magic ‘T’ and has the S-matrix: 0 0 1 1   1 0 0 1 − 1 S= 2  1 1 0 0  1 −1  0 0  As usual the coefficients of the S-matrix can be found by using the unitary condition and mechanical symmetries. strip line.c) double-hole.e) multiple-hole types There is a common principle of operation for all directional couplers: we have two transmission lines (waveguide. i. 76 . such as small matching stubs in the centre of the ‘T’. Provided both generators have equal amplitude and phase. 9. 8 assuming two with equal amplitudes into the E and H arm. The bandwidth of a waveguide generator 2 propagates in the same way. and the sum (Σ) and difference (∆) signals are available from the H arm and E arm. They are widely used for signal combination or splitting in pickups and kickers for particle accelerators. T-hybrids are often produced not in waveguide technology. coaxial line. which correspond to the ∆ and Σ ports. reciprocal.

. The directivity is the ratio of the desired coupled wave to the undesired (i. 9b). Note that the ideal 3 dB coupler (like most directional couplers) often has a π/2 phase Practical numbers for the coupling are 3 dB. Each coupling hole excites waves in both directions but the superposition of the waves coming from all coupling holes leads to a preference for a particular direction. while the magnetic coupling is angle-dependent. A propagating wave in the main line travelling waves in both directions. The following relations hold for an ideal directional coupler with properly chosen reference planes S11 = S22 = S33 = S44 = 0 S21 = S12 = S43 = S34 (2. It can be shown that for α = 30° the electric and magnetic components cancel in one direction and add in the other and we have a directional coupler. crosses. 9 is similar.e. To characterize directional couplers.PARAMETERS The single-hole coupler (Fig. e. αd = directivity [dB] . Each of these currents gives rise to waveguides (also possible with two coaxial lines at an angle α). the coupling appears in the S-matrix as the coefficient S13=S31=S42=S24 with αc = -20 logS13in dB being the coupling attenuation.. In order to get directionality. two important figures are always required.6) S31 = S13 = S42 = S24 S41 = S14 = S32 = S23 77 . The electric coupling is independent of the angle α between the excites electric and magnetic currents in the coupling hole. Optimum directivity is only obtained in a narrow frequency range where the The holes need not be circular. they may be longitudinally or transversely orientated slots. S31 20 log S41 than 20 dB. 6 dB. microstrip) and also lumped element couplers (like transformers). For larger bandwidths. The physical mechanism for the other couplers shown in Fig. For the elements shown in Fig. multiple hole couplers are used. at least For the Bethe coupler the electric coupling does not depend on the angle α between the waveguides. two coupling mechanisms are necessary. the coupling and the directivity. For a backwards coupling towards port 4 these two waves have a phase shift of 180°. 9. and 20 dB with directivities usually better shift between the main line and the coupled line (90° hybrid). 9). so they cancel. wrong direction) coupled wave. 10 dB. takes advantage of the electric and magnetic polarizability of a small (d<<λ) coupling hole.g. Example: the 2-hole. also known as a Bethe coupler. λ/4 coupler For a wave incident at port 1 two waves are excited at the positions of the coupling holes in line 2 (top of Fig.RF ENGINEERING BASIC CONCEPTS : S. For the forward coupling the two waves add up in phase and all the power coupled to line distance of the coupling holes is roughly λ/4. Besides waveguide couplers there exists a family of printed circuit couplers (stripline.e. 2 leaves at port 3. i. etc.. two coupling holes or electric and magnetic coupling.

g.8)   0 0 ± j 1 0 As further examples of 4-ports.7) and for the 3 dB coupler (π/2-hybrid) 0 ±j 0 1 ± j  1 1  = 0 0  2 ±j 0 1 S3dB . 10). 11. (2. 78 . At higher frequencies more waveguide modes can propagate. a network acting as a 2-port at low frequencies will become a 2n-port at higher frequencies (Fig. C ASPERS    1 − S13 ± j S13  2   0 0  1 − S13 2 ± j S13    0 0 S =  ± j S 1 − S13    2 0 0   13  ± j S13 1 − S13    2 0 0 (2. Examples are the stripline coupler shown in Fig. 10: Example of a multiport comprising waveguide ports. with n increasing each time a new waveguide mode starts to propagate. For more general cases. the 90°. for calculation of the scattering properties of waveguide discontinuities. one must keep in mind that a port is assigned to each waveguide or TEM-mode considered. Also a TEM line beyond cutoff is a multiport. 3 dB coupler in Fig. the 4-port circulator and the one-to-three power divider should be mentioned. In certain cases modes below cutoff may be taken into account. the port number increases correspondingly. Since for waveguides the number of propagating modes increases with frequency. There are different technologies for realizing microwave elements such as directional couplers and T-hybrids. 12 and the printed-circuit magic T in Fig.. e. Fig. F. 13. using the S-matrix approach.

With the static capacity per unit length. The space around this conductor is filled with a homogeneous dielectric material.1 Striplines A stripline is a flat conductor between a top and bottom ground plane. Fig.PARAMETERS Fig. curve 1: 3 dB coupler. the static inductance per unit length. curve 3: 10 dB coupler (cascaded 3-dB and 10-dB coupler) [2]. L’.RF ENGINEERING BASIC CONCEPTS : S. 13: Magic T in a printed circuit version [2] 3 Basic properties of striplines. C’. microstrip and slotlines 3. 11: Two-stage stripline directional coupler. This line propagates a pure TEM mode. the relative 79 . 12: 90° 3-dB coupler [2] Fig. curve 2: broadband 5 dB coupler.

14) is given by L′ Z0 = C′ ν ph= = c 1 εr L′C ′ . dimensions and characteristics for offset centre-conductor strip transmission line [5] 80 .1) Z0 = εr ⋅ C ′c 1 Fig. The total capacity is the sum of the principal and fringe capacities Ctot = Cp1 + Cp2 + 2Cf1 + 2Cf 2 . (3. 15) [5]. 14: Characteristic impedance of striplines [5] For a mathematical treatment. the effect of the fringing fields may be described in terms of static capacities (see Fig. F. εr. Cp and Cf. the characteristic impedance Z0 of the line (Fig. C ASPERS permittivity of the dielectric. (3. c. and the speed of light.2) Fig. 15: Design.

right: broad-coupled parallel lines A graphical presentation of Eqs.15 Ω (3. 17. 94.even = Z 0 1 + C0 (3.even where C0 is the voltage coupling ratio of the /4 coupler. for all TEM-modes.3) Z 0. 81 . They are given as a good approximation for the side coupled structure (Fig.even = ⋅ 1 εr w ln 2 1   π s  + + ln1 + tanh   b π π   2 b  94.RF ENGINEERING BASIC CONCEPTS : S.odd = ⋅ 1 εr w ln 2 1   π s  + + ln1 + coth   b π π   2 b  Fig.3) is also known as the Cohn nomographs [5]. left) [5].4) Z 0 = Z 0. Fig. For a quarter- wave directional coupler (single section in Fig. an odd mode and an even mode (Fig.odd Z 0.PARAMETERS For striplines with an homogeneous dielectric the phase velocity is the same. Besides the phase velocity the odd and even mode impedances Z0. 17: Types of coupled striplines [5]: left: side coupled parallel lines. 16). A configuration of two coupled striplines (3-conductor system) may have two independent TEM-modes. 16: Even and odd mode in coupled striplines [5] The analysis of coupled striplines is required for the design of directional couplers. (3.odd = Z 0 1 − C0 1 − C0 Z 0. and frequency independent. They are valid as a good approximation for the structure shown in Fig. 1 + C0 Z 0.15 Ω Z 0.even must be known. 17. 11) very simple design formulae can be given.odd and Z0.

i. 18). 20 and 21 various planar printed transmission lines are depicted. Microstrip circuits are also known as Microwave Integrated Circuits (MICs). 18: Microstripline: a) Mechanical construction. changes in width. The mains reasons are the cheap production. In Figs. etc. Owing to dispersion of the microstrip. Since there is a transversely inhomogeneous dielectric. b) Static field approximation [6] An exact field analysis for this line is rather complicated and there exist a considerable number of books and other publications on the subject [6. and very cheap high-quality elements are available in a wide frequency range. An even simpler way to make such devices is to use a section of shielded 2-wire cable. Stripline coupler technology is now rather widespread. This has several implications such as a frequency-dependent characteristic impedance and a considerable dispersion (Fig. and only part of its cross section is filled with a dielectric material. the calculation of coupled lines and thus the design of couplers and related structures is also more complicated than in the case of the stripline. 19). F. once a conductor pattern has been defined. It is thus an asymmetric open structure.. 3. metallic strip: ρ ρ Fig. 7]. only a quasi-TEM wave exists. The microstrip with overlay is relevant for MMICs and the strip dielectric wave guide is a ‘printed optical fibre’ for millimetre-waves and integrated optics [7]. A further technological step is the MMIC (Monolithic Microwave Integrated Circuit) where active and passive elements are integrated on the same semiconductor substrate. through holes.e. Microstrips tend to radiate at all kind of discontinuities such as bends. C ASPERS In contrast to the 2-hole waveguide coupler this type couples backwards. and easy access to the surface for the integration of active elements. the coupled wave leaves the coupler in the direction opposite to the incoming wave. one may question why they are used at all.2 Microstrip A microstripline may be visualized as a stripline with the top cover and the top dielectric layer taken away (Fig. 82 . With all the above-mentioned disadvantages in mind.

21: Various transmission lines derived in MIC from the microstrip concept [7] 3. A unique feature of the slotline is that it may be combined with microstrip lines on the same substrate. 83 . 19: Characteristic impedance (current/power definition) and effective permittivity of a microstrip line [6] εr εr εr2 εr εr εr2 εr2 εr1 εr εr εr1 Fig. eff Fig. 20: Planar transmission lines used Fig. 22.RF ENGINEERING BASIC CONCEPTS : S. It is essentially a slot in the metallization of a dielectric substrate as shown in Fig. The characteristic impedance and the effective dielectric constant exhibit similar dispersion properties to those of the microstrip line.3 Slotlines The slotline may be considered as the dual structure of the microstrip.PARAMETERS εr.

the signal leaving the circuit on the lower microstrip is inverted since this microstrip ends on the opposite side of the slotline compared to the input. 23: Two microstrip–slotline transitions connected back to back for 180° phase change [7] 84 . Fig. Assuming the upper microstrip to be the input. permits interesting topologies such as pulse inverters in sampling heads (e. d) Magnetic line current model. Figure 23 shows a broadband (decade bandwidth) pulse inverter. Reproduced from Ref... [6] with permission of the author.g. e. for sampling scopes).g. Fig. F. c) Longitudinal and transverse current densities. for stochastic cooling [8]. 22: Slotlines a) Mechanical construction. b) Field pattern (TE approximation). Printed slotlines are also used for broadband pickups in the GHz range. in conjunction with through holes. C ASPERS This.

. one would like to return to S-parameters S= = T11 − 12 21 T12 T T 11 . T12 S21 S21 (A2) T21 = T22− = S22 1 . In the following. for the cascade. 3]: TM = T1 T2  Tm (A3) There is another definition that takes a1 and b1 as independent variables. S12 T22 T22 (A7) S21 = . a matrix multiplication from ‘right’ to ‘left’. (A6) i.  1   21 T22   b2  (A1) As the transmission matrix (T-matrix) simply links the in. S12 S12 Then. T12 S12 S12 (A5) T21 = T22− = S11 1 .  b2   T11 T12   a1    =      a2   T21 T22   b1  (A4) and for this case −T11 = = S22 S11 S22 S21 . S12 = − 21 .and outgoing waves in a way different from the S-matrix. (A1) will be applied. S21 S21 The T-matrix TM of m cascaded 2-ports is given by a matrix multiplication from the ‘left’ to the ‘right’ as in Refs.e. after having carried out the T-matrix multiplication. [2. (A8) 85 . . one may convert the matrix elements mutually −T11 = = S22 S11 S11 S12 .PARAMETERS Appendix A: The T-matrix The T-matrix (transfer matrix). which directly relates the waves on the input and on the output. In practice. 1 T T22 T22 For a reciprocal network (Sij = Sji) the T-parameters have to meet the condition det T = 1 T11T22 − T12T21 = 1. we obtain  =T  T   TM m m−1  T1 .. is defined as [2]  b1   T11 T12   a2   a  = T   . . the definition using Eq.RF ENGINEERING BASIC CONCEPTS : S.

all the other ports have to be open. C ASPERS So far. F. S12 −A+ B − C + D (A13) = = 2 A+ B +C + D A+ B +C + D S12 .  N  1  The elements of the S-matrix are related as A+ B −C − D = = 2 det A A+ B +C + D A+ B +C + D S11 . a Y-matrix (admittance matrix) can be defined as =I1 Y11U1 + Y12U 2 or (= I) (Y ) ⋅ (U ) .and Y-matrices are not easy to apply for cascaded 4-poles (2-ports). we have been discussing the properties of the 2-port mainly in terms of incident and reflected waves a and b. For example. 1 a minus sign appears for I2 of a first 4-pole becomes I1 in the next one. It has dimensionless elements only and is obtained by dividing B by Z0 the reference impedance. A description in voltages and currents is also useful in many cases. Furthermore. in contrast to the S-parameter measurement. the impedance Z (Fig. 4] A B  1 Z/Z 0  C D = 0 . the so-called ABCD matrix (or A matrix) has been introduced as a suitable cascaded network description in terms of voltages and currents (Fig. respectively. (A10) = I 2 Y21U1 + Y22U 2 Similarly to the S-matrix. Considering the current I1 and I2 as independent variables. In an analogous manner. = U 2 Z 21I1 + Z 22 I 2 (A9) where Z11 and Z22 are the input and output impedance. 1) with ZG = ZL = Z0 would have the normalized ABCD matrix [3. S22 . When measuring Z11. the H matrix (hybrid) and G (inverse hybrid) will be mentioned as they are very useful for certain 2-port interconnections [3]. the dependent variables U1 and U2 are written as a Z matrix: = U1 Z11I1 + Z12 I 2 or (U = ) (Z ) ⋅(I ) . It can be shown that the ABCD-matrix of two or more cascaded 4-poles becomes the matrix product of the individual ABCD-matrices [3]: A B A B A B A B = ⋅⋅⋅ (A12) C D C D C D C D . and multiplying C with Z0. = U1 H11I1 + H12U 2  U1   I1   I=  (H ) ⋅   2 U 2  or (A14) =I 2 H 21I1 + H 22U 2 86 . to the elements normalized of the ABCD matrix.  1   2   21 A22   − I 2  With the direction of I2 chosen in Fig. 1):  U1   A B   U 2   A11 A12   U 2  = = (A11)  I  C D   −I  A  .  K  1  2  k In practice. Thus. the Z. where matched loads are required. the normalized ABCD matrix is usually applied.

A transmission line section     Ch − Sh    1     0 Sh ( )  Sh  Ds    2  2ZZ0 2ZZ0 γl  Ch   2ZZ Z2 − Z02 Sh   − Z − Z0 Sh Ch + Z + Z0  Z  2 2 2    2ZZ0    0 Sh 2ZZ0 where Sh = sinhγ.RF ENGINEERING BASIC CONCEPTS : S. S.PARAMETERS and =I1 G11U1 + G12 I 2  I1  U1  or  =  (G ) ⋅   . one may work with S-matrices directly.1: Basic interconnections of 2-ports [1]: a) Parallel–parallel connection add Y-matrix. d) Parallel–series connection add G matrix [3]. the four basic possibilities of interconnecting 2-ports (besides the cascade) are shown. Figure A. in computer codes for RF and microwave network evaluation.1. eliminating the unknown waves at the connecting points by rearranging the S-parameter equations. Fig. In simple cases.2 shows ABCD-. b) Series–series connection add Z-matrix. A.and T-matrices (reproduced with the permission of the publisher [3]). in particular. ( ) Element ABCD matrix S matrix T matrix Ch Z Sh   Z2 − Z2 Sh 2ZZ0   Z2 + Z02 Z2 − Z02  1. in Fig. but they are applied. As an example. c) Series–parallel connection add H matrix. U 2   I2  (A15) = U1 G12U1 + G22 I 2 All these different matrix forms may appear rather confusing. A. Ch = coshγ and Ds = 2ZZ0 Ch + (Z2 + Z02) Sh 87 .

C ASPERS 1 Z  Z + Z2 − Z1 2 Z1Z2   Z1 + Z2 − Z Z2 − Z1 + Z    1   1   2. A shunt     Dt   admittance  Y 1 Ds    Y1 − Y2 + Y Y1 + Y2 + Y   2 Y1Y2 Y2 − Y1 − Y  2 Y1Y2 where Ds = Y + Y1 + Y2 and Dt =  −1 0   −1 Ds − 1  Z0 Z0   1  1   1 − 2Z T −j 2Z  4. An ideal       transformer 1 0 1 / n  n +1  2 2n   1− n  n − 1 n + 1 2  2n 2 2 7. Z0  jT      j connected short-  1 Ds    − j Z0 Z0  Z  1− j  Ds + 1 circuited stub 1   2ZT 2ZT  where T = tanβ and Ds = -1 + 2jZ/(Z0T) n 0  n 2 − 1 2n  n 2 + 1 n 2 − 1     1   6. D = Y1Y2 + Y2Y3 + Y3Y1. A series     Dt   impedance 0 1  Ds    2 Z1Z2 Z + Z1 − Z2   Z2 − Z1 − Z Z1 + Z2 − Z  2 Z1Z2 where Ds = Z + Z1 + Z2 and Dt = 1 0  Y1 − Y2 − Y 2Y1Y2   Y1 + Y2 − Y Y1 − Y2 − Y    1   1   3. F. A shunt. Q = Y1 + Y2 + 2Y3 and P = Y1 – Y2 88 . A shunt-  1     T connected open- Ds    jZT   j Z0 T Z0  1+ j T  Ds − 1 − 1  ended stub  2Z 2Z  where T = tanβ and Ds = 1 + 2jZT/Z0 1 0 1 Ds + 1  Z0    1   1 + j 2ZT 2ZT  5. π-network  Y2 1   Y02 − PY0 − D 2Y0 Y3   −Y02 + QY0 − D Y 2 − PY0 − D  1+ Y  1       3   1   0 Ds   2Y0 Y3   Y D Y  3    −Y02 − PY0 + D Y02 + QY0 + D   1+ 1   2Y0 Y3 Y02 + PY0 − D     Y3 Y3  where Ds = Y02 + QY0 + D.

S.and T-matrices for the elements shown Appendix B: The signal flow graph (SFG) The SFG is a graphical representation of a system of linear equations having the general form y = Mx + M y (B1) where M and M are square matrices with n rows and columns. each arrow stands for an S-parameter (Fig. B.1).PARAMETERS  Z1 D   − Z02 + PZ0 + D 2Z0 Z3   − Z02 + QZ0 − D − Z02 + PZ0 + D  1+ Z    1   8. Each wave ai and bi is represented by a node. x represents the n independent variables (sources) and y the n dependent variables. A transmission       Ds   line junction Ds    Z2 − Z1 Z1 + Z2  0 1  2 Z1Z2 Z1 − Z2  2 Z1Z2 where Ds = Z1 + Z2 and Dt = α-db A + B  A − B  0 B  −A 0  2 Z0       10. B. as is generally the case in practice.1: A 2-port with a non-matched load 89 . Q = Z1 + Z2 + 2Z3 and P = Z1 – Z2 1 0  Z2 − Z1 2 Z1Z2   Z1 + Z2 Z2 − Z1    1   1   9. a1 Fig. the previous equation simplifies to y = Mx. T-network  3  1     Z  1 Z  Ds   2Z0 Z3   3  1+ 2     Z0 + PZ0 − D Z02 + QZ0 + D   3 Z3   0 3 − − +  2 2 Z 2Z Z Z 0 PZ 0 D where Ds = Z02 + QZ0 + D. The elements of M and M appear as transmission coefficients of the signal path. A. RF ENGINEERING BASIC CONCEPTS : S. An attenuator   2      A -B A+B         B 0  0 A   0  2Z 2A where A = 10α/20 and B = 1/A Fig.2: (continued) ABCD-. D = Z1Z2 + Z2Z3 + Z3Z1. which is equivalent to the usual S-parameter definition b = Sa . (B2) The SFG can be drawn as a directed graph. When there are no direct signal loops.

C ASPERS For general problems the SFG can be solved by applying Mason’s rule. No signal paths should be interrupted when resolving loops.1) As before. since loops can be transformed to forward and backward oriented branches. The SFG is also a useful way to gain 90 . a2 is added to S11. B. given a branch from b2 to a2 with the = S11 + S21 b1 1 − S22 L S12 . a more intuitive way is to simplify step-by-step the SFG by applying the following three rules (Fig. We are looking for the input reflection coefficient b1/a1 of a two-port with a non-matched load ρL and Examples a matched generator (source ρS = 0). the solution may sometimes be read directly from the diagram. yielding a loop with ρS. Multiply the signals of cascaded branches 3. Add the signal of parallel branches 2. B. Finally one obtains    S 11 + S 21 S 12 ρ L  1 − ρ L S 22 1   . Then the signal path via b2 and and ask for b1/bs. F. Applying the cascading rule and the parallel branch rule then yields The loop at port 2 involving S22 and L can be resolved. When applied to S-matrices. parallel 2. Resolve loops x x x y y y x+y xy x/(1-xy) 1. B.2): 1. 10]. the SFG is a convenient tool for the analysis of simple circuits [9. cascaded 3. first the loop consisting of S22 and ρL can be resolved. loops branches signal paths Fig. As we have seen in this rather easy configuration. see Fig. For not too complicated circuits.1. (B3) a1 L As a more complicated example one may add a mismatch to the source (ρS = dashed line in Fig. For more complex networks there is a considerable risk that a signal path may be overlooked and the analysis soon becomes complicated. B. = b1   (B4) 1 −  S 11 + S 21 S 12 ρ L  ρ S bS 1 − ρ L S 22 1   The same results would have been found applying Mason’s rule on this problem. signal ΓL *(1-ΓL*S22).2: The three rules for simplifying signal flow charts Care has to be taken applying the third rule.

RF ENGINEERING BASIC CONCEPTS : S.PARAMETERS insight into other networks. such as feedback systems. B.3: SFG and S matrices of different multiports (reproduced from Ref. the need to use the SFG has been reduced. [10] with the permission of the publisher) The purpose of the SFG is to visualize physical relations and to provide a solution algorithm of Eq. Element S matrix SFG b = ρia a) Passive one-port b = bq + ri a b) Active one-port  S11 S12  S =   S21 S22  c) Passive two-port 0 e− j l  S =   e− j 0   l d) Lossless line (length ) matched  S11 S12 S13    S =  S21 S22 S23  S   31 S32 S33  e) Passive 3-port  S11 S12 ⋅⋅⋅ S1n     S21 S22 ⋅⋅⋅ S2n  S =  ⋅⋅ ⋅⋅ ⋅⋅⋅ ⋅⋅     ⋅⋅ ⋅⋅ ⋅⋅⋅ ⋅⋅  S   n1 Sn 2 ⋅⋅⋅ Snn  f) Passive n-port Fig. (B4) by applying a few rather simple rules: 91 . But with the availability of powerful computer codes using the matrix formulations.

starting from a node and going along the arrows back to that node without touching the same node more than once. going from N1 to N2. C ASPERS 1. each representing one S-parameter and indicating direction. branch (arrow). N1 has all signals (branches) previously entering N. 4. The transmission coefficients of parallel signal paths are to be added. Examples We are looking for the input reflection coefficient of a e-port with a non-matched load ρL and a matched generator (source) (ρS = 0) to start with. 2. etc.  ΣL(1)(1) represents the sum of all first-order loops not touching path 1. ρS are often written as ΓL. in the denominator are the sums of all first-. 8. and N2 is linked to all signals previously leaving from N.4: 2-port with non-matched load 92 . Fig. A node may be the beginning or the end of a All other nodes are dependent signal nodes. 3. An elementary loop with the transmission coefficient S beginning and ending at a node N may be replaced by a branch (1-S)-1 between two nodes N1 and N2. A second-order loop is the product of two non-touching first-order loops. second-. 6. In order to determined the ratio T of a dependent to an independent variable the so-called ‘non- touching loop rule’. The SFG has a number of points (nodes) each representing a single wave ai or bi. 5. Each node signal represents the sum of the signals carried by all branches entering it. F. 9. also known as Mason’s rule. 7. 10.  The expressions ΣL(1). ΓS. A first-order loop is the product of branch transmissions. may be applied [11] P1 1 − ∑ L (1)( ) + ∑ L ( 2 )( ) − ⋅⋅⋅ + P2 1 − ∑ L (1)( ) ⋅⋅⋅     1 1 2 T= 1 − ∑ L (1) + ∑ L ( 2 ) − ∑ L ( 3) + ⋅⋅⋅ (B5) where:  Pn are the different signal paths between the source and the dependent variable. Nodes are connected by branches (arrows). and ΣL(2)(1) is the sum of all second-order loop not touching path 1.  Analogously ΣL(1)(2) is the sum of all first-order loops in path 2. The transmission coefficients of cascaded signal paths are to be multiplied. Nodes showing no branches pointing towards them are source nodes. ρL. An SFG is feedback-loop free if a numbering of all nodes can be found such that every branch points from a node of lower number towards one of higher number. and an nth-order loop is the product of any n non-touching first-order loops. order loops in the network considered. ΣL(2) etc. B.

Heidelberg. Heidelberg.C. Garg. Chadha. [11] S. Bahl.). 1979). Bryant and S. P. 93 . Elektronik Anzeiger No. vol. Howe.RF ENGINEERING BASIC CONCEPTS : S. 1986).L. Gundlach. Garg. 1991. Yip. B.C. (B8) S L L S21 S 22S22 L References [1] K. [2] H. [9] P. J. Mikrowellenschaltung (Springer. Proc. London.J. and I. 11. Passive nichtreziporke Bauelemente der Mikrowellentechnik. B. Fiebig. Microwave kickers and pickups. London. CERN Accelerator School: RF Engineering for Particle Accelerators. [10] A.-P. 1980). CERN 85–07 (1985). (also available in English). Entschladen. Frascati. Auflage (Springer. New York. Dupont. [4] H. Columbia. MA. Microstrip Lines and Slotlines (Artech. Hoffmann. 2. Dedham. 458–471. 217–223. ECFA 85/91. Archiv der Elektrischen Übertragung (AEÜ) 15 (1961) 285–292. Gupta. Power waves and the scattering matrix. [7] K. MA. Microwave Theory Tech. Lineare Signalflussdiagramme. Meinke and F. Mason. Kurokawa. Taylor. MA. [3] K. Taschenbuch der Hochfrequenztechnik. The Generation of High Fields. Stripline Circuit Design. Dedham.K. CERN 92-03 vol.C. Turner (Ed. on Radiation Hazards. [5] H. Feedback theory – Some properties of signal flow graphs. 1981). Integrierte. P. and R.5) b1 S11 (1 − S22 L ) + S21 = L S12 1 − S22 L . [8] C.C. Oxford. 13 (1965) 194–202. 1961). 2. Newman (Eds. Conf. 1984. CAS– ECFA–INFN Workshop. Computer-Aided Design of Microwave Circuits (Artech. R. 4. High-Frequency Circuit Design and Measurements (Chapman and Hall. pp. Dedham. Proc. Gupta.4) and ask for b1/bs S11 (1 − S22 ) + S21 = bs 1 − ( S11 ) + S11 b1 S S12 + S22 + S12 S . (B. High- Frequency Circuit Design and Measurements (Chapman and Hall. p. 1982). IRE 41 (1953) 1144–1156. 1971.4) we obtain = S11 + S21 b1 1 − S22 L S12 a1 L (B6) or by formally applying Mason’s rule in Eq.). S. 1983). (B7) a1 As a more complicated example one may add a mismatch to the source (ρS = dashed line in Fig. Yip. 4th Edition (Artech.L. 1980). R. 1960 (Academic Press.PARAMETERS By reading directly from the SFG (Fig. 396. Int. [6] R. IEEE Trans.